Clinical data warehouse boosts research, patient care quality

At least for larger facilities, the answer to the problem of hospital patient data in discrete silos could be a clinical data warehouse, as the UNC Healthcare System discovered.

One way for large hospitals to satisfy the slate of more than 100 patient-care quality indicators that’s currently...

proposed in federal meaningful use rules, one health IT professional believes, is with a clinical data warehouse.

These repositories, also known as data marts, are served by middleware that doles out data on demand as individual applications make queries. They can make research go much more quickly than in the typical hospital-network IT scenario -- discrete data silos holding different pieces of electronic health record (EHR) systems, or full patient records housed in separate, noninteroperable systems that need to be combined for effective research.

Donald Spencer, associate director of medical informatics at the University of North Carolina (UNC) Healthcare System in Chapel Hill, N.C., said he’s convinced his campus’s new clinical data mart, dubbed the Carolina Data Warehouse for Health, is making patient care more efficient and quality outcomes easier to assess for clinical analysts. It is also helping academics on the university side better aggregate data for grant-funded clinical research studies.

“I really don’t see how an institution our size can achieve meaningful use without a data warehouse of one kind or another, because if you look at the meaningful use requirements, they’re asking for information across patients, population information,” said Spencer, who is speaking about the UNC clinical data warehouse at this year’s Health Information and Management Systems Society (HIMSS) conference.

“How do you do that, unless you have a separate warehouse, a query-able database that incorporates multiple systems?” Spencer said. “You can’t do it. When I was in a solo practice in a town of 300, … the answer to that was, I’d use an IBM PC and run the reports at night. But this is a 24/7 operation, and you can’t run all those population type reports without a separate … optimized database.”

Improved data warehouse access means speedier results

UNC’s clinical data warehouse runs on IBM databases and middleware, and brings together software systems from many other vendors throughout several of the health system’s facilities, including the state-run UNC Medical Center, clinics, physician practices and other affiliated providers. During the past two decades, UNC Healthcare installed systems for billing, computerized physician order entry and EHRs.

When hospital clinicians and UNC researchers previously queried data for quality improvement or research studies across those systems, the laborious process slowed the operations side of the network. So it had to be done during slow periods -- the middle of the night, for example -- so as not to interfere with doctors and nurses caring for patients.

“Because it’s optimized for searching, as opposed to editing and deleting and updating, you can search millions of rows very quickly,” Spencer said of the improved data warehouse access. “It takes much longer to formulate your question than to run your query.”

With the clinical data warehouse, researchers have access to diagnosis and treatment data via a Web portal, and can access more data, more quickly, without slowing the health care providers also accessing data via the EHR system. In the year since UNC installed the data mart, Spencer has helped make more than 100 requests on the academic side for researchers working on grants. That wouldn’t have been possible before, he said.

On the patient-care side, UNC still is establishing metrics for measuring improvements, as analysts explore the possibilities of their newfound access to clinical statistics. One focus is achieving certification from the Washington, D.C.-based National Committee for Quality Assurance (NCQA) certification for diabetes care for some of its clinics. The clinical data warehouse enables the aggregation of diabetes patients’ blood-pressure statistics, a key quality indicator. It helps the health system measure its progress and the effects of its quality initiatives on the bottom line.

“We can go back to the administration and say we improved our health insurance reimbursements,” Spencer noted. “That’s part of the agreement with some of our carriers. Our reimbursement is based on performance, and we have to be NCQA-certified.”

In treatment areas where national standards are not as prevalent or regional trends require more research on UNC’s population, the clinical data warehouse will offer more avenues of research exploration, Spencer said. One such area could be stroke treatment: The Carolinas sit in the so-called stroke belt, where inhabitants have a one-third greater risk of suffering a stroke than the general U.S. population.

Data warehouse security, governance are key issues

As with many initiatives involving EHR systems and patient data, software and middleware might provide a toolkit that enables better workflow, but the human component of governance is a large part of making it work. UNC established a committee to determine rules for who can use the clinical data warehouse and for what purposes, Spencer said.

I really don’t see how an institution our size can achieve meaningful use without a [clinical] data warehouse of one kind or another.
Donald Spencer associate director of medical informatics, UNC Healthcare System

Spencer serves as a sort of traffic cop to make sure that, for instance, academic researchers from the hospital and the schools involved get only the data they need for the specific projects on which they are working. In some cases, that includes making sure patient data is de-identified (all HIPAA-protected identifiers are stripped from the records) or that patient consents, as required by law and UNC’s research institutional review board, are in place, or both. The committee also determines each stakeholder’s share in maintaining and funding the clinical data warehouse.

“The governance really predates the data warehouse,” Spencer said. “You can’t stand up the warehouse, and you can’t continue it after the consultants leave without [the governance committee]. We haven’t had big controversies, but we certainly have had big projects that we haven’t been able to fund yet because they’re so big.”

For other facilities planning a clinical data warehouse, a governance committee is the engine that makes it go, Spencer said. Make sure that all stakeholders are fairly represented on the committee and that it fosters an environment where everyone gets their say -- yet keeps the health system’s goals ahead of any particular individual’s, he advised.

The data mart stores the whole of UNC’s patient data, including HIPAA-protected information. In the past, researchers might have accessed patient data via laptop or hard drive. That transfer method left data vulnerable to its handler, who might lose track of the hardware and cause a data breach. Now, the data stays in the warehouse, securely accessed via a Web portal.

“[Users] sign a data access agreement; they can’t remove the data from the secure environment; if they want to use a statistical tool or something, they can use it on the data, but they can’t physically remove the data,” Spencer said of UNC’s clinical data warehouse security. “We have much more control ... particularly with the governance layer, we know what’s going on more [than before]. The security, in that sense, has more administrative control.”

Let us know what you think about the story; email Don Fluckinger, Features Writer.

More on improving access to health information

  • Health care IT infrastructure burdened by new technology
  • SaaS-based clinical data analytics seen as next wave of automation

Dig Deeper on Clinical data analytics software and systems

Cloud Computing
Mobile Computing